Understanding AI in Education

Exploring the environmental impacts of artificial intelligence

We firmly believe that artificial intelligence can transform education for the better. But we are also aware that for AI to be a force for good, it must be used in a responsible way. The environmental impacts of AI are hence of paramount importance. 

In this blog, we introduce some of the key issues around the environmental impacts of AI. We will also publish follow-up blogs, diving more deeply into specific issues on this topic. 

Why is it important to know about the environmental impacts of AI? 

 In short, because the stakes are huge.  

On the one hand, AI promises to reduce carbon emissions by increasing the efficiency with which societies and industries make use of resources. 

Then again, AI systems themselves can lead to significant carbon emissions. An oft-cited statistic that causes particular alarm is that the environmental cost of training some of the more advanced AI systems is equivalent to 125 return flights from New York to Beijing. This statistic is particularly relevant to the education sector, as some of the most energy intensive Natural Language Processing models are nowfOUNDAT being used to automatically generate written content, including students’ essays. 

How can AI have a positive impact on the environment? 

According to PwC, anticipated applications of AI in energy, water, transport and agriculture could lead to a 4% reduction in greenhouse gas emissions by 2030. Here, environmentally beneficial use cases include using AI to help optimise how land is used for farming and forestry, and developing more efficient methods of delivering goods, which make use of autonomous vehicles. 

The smart city is another application of AI that could have positive environmental impacts. By using data gathered from Internet of Things devices, urban environments are likely to become increasingly flexible, adapting to footfall, traffic flows and energy demands, for instance. This, again, could spearhead efficiencies and lessen cities’ environmental footprints. 

As cities become smarter, so too will campuses. Data-driven insights are likely to help colleges and universities to make more efficient use of their resources – from physical spaces to cloud networks. 

In the construction industry it has been forecasted that AI could enable cement to be produced more efficiently. When considering that cement production is responsible for around 8% of carbon emissions, it becomes clear that relatively small efficiencies could have a big impact globally. This may become particularly relevant for education institutions, who are likely to come under increasing pressure to build new facilities as volumes of students increase. In a recent blog, we looked at the environmental implications of scaling education, and provided some forecasts of the carbon footprints of new buildings. 

Furthermore, as discussed in Jisc’s exploring digital carbon footprints report, tools such as The UK’s Carbon Intensity API use machine learning to monitor and forecast technologies’ contributions to carbon footprints: insights that could help encourage people and institutions adapt and use systems in a more sustainable way. 

How can AI have a negative impact on the environment? 

Put simply, AI systems consume massive amounts of energy, particularly during the training phase. Naturally, this energy usage comes at a price.  

As stated in an article in The New Scientist 

“Training artificial intelligence is an energy intensive process. New estimates suggest that the carbon footprint of training a single AI is as much as 284 tonnes of carbon dioxide equivalent – five times the lifetime emissions of an average car.” 

What’s more, the average amount of energy used in developing AI systems is increasing. One estimate is that this energy expenditure has been doubling every 3.4 months for around 10 years 

One reason for this rate of change is the increasing complexity of AI systems. Take GPT-3, an AI model that is raising questions about the future of assessment due to its ability to generate high quality written content.  GPT-3’s training involved 175 billion parameters, 100 times more than GPT-3’s predecessor, GPT-2. This increase in complexity has contributed to GPT-3’s gains in performance, which in turn has resulted in some convincing and well-written outputs (you can find out more about GPT-3 in our blog on the subject).   

 As colleges and universities consider how to respond to such advances in AI, they will need to consider not only implications for pedagogy and assessment: they will need to consider the environmental impacts too.  

What can be done to reduce the negative environmental impacts? 

If AI systems are to continue to become more accurate without these adverse environmental impacts being sustained, then new, more energy efficient ways of training AI systems will need to be developed. 

‘Green AI’ is thus a growing field of research, the outcomes of which will help determine how much the planet can benefit from AI in the long run. 

Part of the solution for developing green AI will be more efficient hardware. Continuing efforts will be needed to optimize chips, such as graphics processing units and tensor processing units.  

Failing solutions that yield highly accurate, energy efficient AI systems, a further option moving forward will be to finely balance the trade-offs between accuracy and environmental impacts. Cost-benefit analyses could become commonplace, so too could pressure from customers and wider society for the negative environmental impacts of AI systems to be mitigated as far as possible. 

What next? 

It is not hyperbolic to state that finding greener ways to develop and deploy AI will be amongst the most important of all human enterprises. And given the scale of the challenge, all the answers are not yet known.  

That said, a key first step is to become actively mindful of the environmental footprints of different AI software. Mirroring Jisc’s Exploring digital carbon footprints report, it is important to “monitor, measure, and communicate” the carbon footprints of AI systems that are in use, whilst also taking environmental considerations into account when procuring or outsourcing. We’ll be producing practical guidance to help with this over the next few months. 


Find out more by visiting our National centre for AI page to view publications and resources, join us for events and discover what AI has to offer through our range of interactive online demos.

For regular updates from the NCAI sign up to our mailing list.

Get in touch with team directly at

By Tom Moule

Product Lead at The National Centre for AI in Tertiary Education

Leave a Reply

Your email address will not be published. Required fields are marked *